Evolutionary Audio - Keith

kenh@snook-eu
EVOLUTIONARY AUDIO
SOLI D-STAlE
AUDIO POWER
V
alve audio amplifiers generally
require an output coupling
transformer to match the output
impedance to that of the loud­
speaker load. If a good performance is
sought, this component will be expen­
sive and bulky. The savings in cost and
bulk which are possible and the im­
provement in performance, especially
at the extreme ends of the audio spec­
trum, by avoiding the need for this
component have remained one of the
major benefits of "solid-state" circuitry.
Lin circuit, which was designed around
germanium transistors, gave misleading
encouragement to other engineers, on a
world-wide basis, who translated the
design into silicon-transistor-based ver­
sions when, during the early 1960s,
n-p-n silicon planar power transistors
became available.
The inherent snag in this approach is
that the base voltage/collector current
characteristics of germanium and sili­
con transistors are different, with that of
the silicon device being much more
abrupt, as shown in Fig. 2.
Moreover, since the permissible ther­
mal dissipations of the output devices
were then fairly limited - by comparison
with valves - it was necessary to operate
the output stages at a fairly low quies­
cent current, in class AB, or even (with
Early transistor audio
amplifiers
Understandably, early designs in this
field owed a lot to previous valve ampli­
fier practice, with transformer inter­
stage coupling being used to allow a
push-pull output configuration. How­
ever, the real break-through in this field
came with the introduction, in 1956, of
the " quasi-complementary" output
stage due to H. C. Lint, of which the
basic circuit layout is shown in Fig. 1.
At the time, the most easily obtained
transistors were germanium diffused­
junction p-n-p devices, although some
germanium n-p-n transistors were be­
coming available in low-power versions.
The attractive feature of Lin's design
was that the circuit provided a push-pull
output without the need for a coupling
transformer, and with a pair of output
transistors which were both of the same
type. In addition, it only required one
low-power n-p-n device.
The performance of this circuit was
excellent by contemporary transistor
audio-amplifier standards, in that it had
a 30Hz - 15kHz bandwidth and a full­
output-power THD figure of less than
1% at 1kHz, which decreased some­
what with decreasing output ;>ower.
However, germanium transistors have
too high a temperature coefficient of
leakage current for them to be suitable
for domestic use, where thermal run­
away could never be completely ruled
out.
Sadly, the relative excellence of the
1042
John Linsley Hood
traces the evolution of
transistor audio
power amplifiers
from 1956 to the
present day. Designs
produced up to 1975,
covered in this first
part, reached a high
standard, but stir!
contained residual
design mistakes
zero quiescent current) in class B. High
(notional) levels of negative feedback
were then used to lessen the residual
distortion which this incurred.
This design philosophy had the unfor­
tunate effect of maximizing the per­
formance penalties, in that the high
levels of NFB inevitably contributed to
poor overall loop-stability margins
while, at the "crossover" point, the
effective gain of the output devices was
low or even zero, so that the NFB was
ineffective in reducing the distortion at
the very point where it would have been
useful.
Also, because of the basic asymmetry
of the "quasi-complementary" output
stage, as shown in Fig. 3, not only was
the residual inherent distortion large,
but it tended to increase as the output
power level was reduced, as shown in
Fig. 4.
This meant that a manufacturer's
specification which claimed, for exam­
ple, "better than 0. 05% THD at full
output power" might be quite irrelevant
to the user, who might have to put up
with ten times this amount of distortion
at his normal listening levels.
Moreover, the residual distortion,
especially at low powers, was rich in
dissonant harmonics, which were alien
to the normal experience of the human
ELECTRONICS WORLD + WIRELESS WORLD
November 1989
EVOLUTIONARY AUDIO
-Ver
ear. In addition, the reduced gain at the
point at which the signal waveform
crossed the zero axis tended to suppress
low-level signal components and give
the amplifier a "thin" sound, lacking in
"warmth" and "richness".
It was hardly surprising, therefore,
that these early silicon-transistor quasi­
complementary "high-fidelity" designs
won few friends among their users.
More regrettably in the long term, this
unfortunate and temporary lapse of
design standards has led to two break­
away movements among the 'hi-fi' com­
munity: the " all specifications are
meaningless, so only believe your ears"
fraternity, and the "back to valves"
brigade.
1000;J
E in
----ll l--�
10�
2N158
LS (16fll
Improved output-stage
configurations
OV
Fig. 1. Original quasi-complementary
transistor power amplifier by H. C. Lin,
from 1956.
Fig. 2. Comparison of silicon and
germanium base-voltage/collector­
current curves at 25°C.
Out
There were, in the 1960s, three practic­
able options for improving the perform­
ance of audio-output stages: to use fully
complementary output devices, which
were just becoming commercially avail­
able; to use the output devices in class
A ; o r t o m o d i f y t h e qu a s i ­
complementary arrangement so that it
gave greater symmetry in the two
halves.
The first of these approaches was
adopted, soon after suitable devices
became available, by Locanthi2 and
Bailey3. The output stages of a 30W per
channel design due to Bailey are shown
in Fig. 5.
There are two difficulties inherent in
this approach, of which the first is that
the p-n-p output devices were, at that
time - and to some extent even today -
I.J
:2:
0
:l:
ex
<t
:x:
....
<t
le
1·2
0,'
�
0
�
0·01
,/
15W Valve
---
--
_
-
-
1·0 1-2
/
------...-/
10
0,'
POWER OUTPUT (WATTS)
0·8 04
/
I
I
I
Overload
1-4 ·VB
Fig. 3. Asymmetry of silicon quasi-complementary pair. Small
diagram shows crossover characteristic when pair optimally
biased.
-le
November 1989
ELECTRONICS WORLD + WIRELESS WORLD
Fig. 4. Asymmetry of early silicon quasi-complementary
amplifiers shown in Fig. 3 gave rise to increasing cross­
over distortion at low power levels, in contrast with the
behaviour of a good-quality valve amplifier.
1043
EVOLUTIONARY AUDIO
--------,
+30V
3k9
Eio --+-.-
�
40362
Fig. 5. Bailey's 30W
amplifier, which used a fully
complementary output stage.
100
J"
3kL
2S}J
NFB
etc
-�1�k��-+----+--���--�'-�-4�-----+�------�-0-R
5
01
.
10
LS
Iov
SOp
1k
3k9
1k
-30V
________
' -.J
rather more fragile than their nominal
n-p-n equivalents, which prompted
Bailey to evolve an effective overload
protection circuit, also shown in the
diagram.
The second problem is that, because
of the different majority carriers in the
two transistor forms, p-n-p devices tend
to have a lower HF transition frequency
than equivalent n-p-n ones. The differ­
ence in the transition frequencies of the
"complementary" output transistors
leads to asymmetry of the output stage
at higher audio frequencies, with a
consequent worsening of crossover and
other distortion characteristics.
At that time my own preference,
provided that the power requirement
was relatively modest, was for the use of
class A operation, and a circuit for a
IOW power amplifier using this
philosophl is shown in Fig. 6. This is
not a push-pull system, and is therefore
intrinsically free from crossover prob­
lems. This particular circuit can be
visualized either as a simple transistor
gain stage with an active collector load,
or as an emitter follower with an active
emitter load. A difficulty in the use of
this layout is that it has a low overall
efficiency and is not easily extended in
power without the use of a bridge
..:onfiguration.
The third approach is exemplified bl
a neat circuit adaptation due to Shaw ,
in which an added diode is used to
lessen the differences between the up-
1044
�-----�------V=cc+27V
11·2A)
r-
___________
MJ480
(2N30ss)
Etu
--I
5000
10k
��
��
__
��
__
��
______
per and lower halves of the output pair,
as shown in Fig. 7(a). Because the
output transistors can then be of identic­
al type (and Ft), the worsening of THD
with increase in frequency can be les­
sened.
Baxandall, following an analysis of
this problem6, suggested an elegant cir­
cuit improvement, shown in Fig. 7(b),
which almost completely eliminates the
dissimilarity between the upper and
lower halves of the output stage, and
allows a low-distortion design to be
made with identical output transistor
types.
For a subse � uent higher-power
amplifier design , I followed in the
____
-A
____
�
LS(80)
"___ OV
__
Fig. 6. Author's 1969 10W class A
amplifier. Since the operation is not
push-pull, there is no crossover
distortion.
footsteps of Shaw and Baxandall, with
the circuit layout shown in Fig. 7(c), in
which I had added a small capacitor to
the resistor/diode network to simulate
the effect of the output transistor base/
emitter capacitance.
An alternative arrangement, intro­
duced commercially by the Acoustical
Manufacturing Co.s in their Quad 303
power amplifier, employed a pair of
quasi-complementary triplets, of the
type shown in Fig. 8. This generates a
ELECTRONICS WORLD + WIRELESS WORLD
November 1989
EVOLUTIONARY AUDIO
--�----------------,-------��------------
---
+Vcc
lkS
NFB etc.
33k
E
in------f�
lS
1
22n
330
TlP30(
2N3442
100
100
-Vdd
+67V
22k
Bootstrop
lk
2N3055
4k7
OR3
15920
x4
-----.....
....____+-_I__
��--E in
-------4�---H
I
��00}J
0
l5
2N3055
��
---____________
November 1989
�
________
________
ELECTRONICS WORLD + WIRELESS WORLD
_a
____
_a_OV
Fig. 7. Shaw's improved quasi­
complementary design from 1969,
which used a diode to improve
symmetry, is seen at (a). At (b),
Baxandall's variation further improves
symmetry, and (c) shows author's use of
small capacitor to simulate effect of
base/emitter capacitance.
high internal loop gain within each of
the compound output emitter-follower
groups, which helps to minimize the
asymmetry of the output stage "halves"
and the residual crossover distortion
which this asymmetry introduces.
Other layouts have been proposed to
improve symmetry in such quasi­
complementary pairs, such as that due
to Visch9 and Stevens 10, but contempor­
ary high-quality design appears to be
exclusively committed to symmetrical
layout employing using complementary
transistors, which use either the output
transistor configuration shown in Fig. 5,
or that of a symmetrical compound
emitter follower of the type shown in
Fig. 9. This has the advantage that the
base/emitter j unctions of the output
devices, which will get hot, are not
included in that part of the circuit which
determines their forward bias, which
offers better output-stage quiescent cur­
rent stability.
All of these class AB circuit layouts
require that the quiescent current in the
output stage remains close to some
optimum value if the target perform­
ance of the design is to be achieved, in
spite of changes in the temperature and
age of the components. This has been
the subject of considerable circuit de­
velopment, of which some radical
approaches are discussed later.
With an eye on their use as output
devices, several manufacturers have in­
troduced low-cost, high-specification,
monolithic, Darlington-connected out­
put transistors, having the internal
structure shown in Fig. 10. However,
because the driver transistor is on the
same chip as the output device and is
heated by it, the use of such output
transistors makes output-stage
quiescent-current stability more diffi­
cult to achieve.
Direct-coupled layouts
All of the earlier "transformerless"
transistor power amplifier layouts were
designed to operate between the av rail
and some single positive (or negative)
supply line, with a DC blocking capaci­
tor to the loudspeaker, using a layout
similar to that shown in Fig. 6. This
Fig. 8. Quad 303 quasi-complementary
triplets.
1045
EVOLUTIONARY AUDIO
+Vcc
etc.
..
t
III
...
.!!
"i:
Output
�
I!
g
...
etc. -....-+--1
----��--��-ov
meant that the loudspeaker unit was
protected from damage in the event of a
semiconductor failure, but involved the
use of a large-value coupling capacitor if
an extended low-frequency response
was sought.
However, designers became in­
creasingly convinced that there were
advantages in sound quality to be
obtained by the use of the so-called
direct-coupled layout, of the type
shown in Fig. 11, in which the amplifier
operated between a pair of symmetrical
(±) supply lines, so that there was no
longer a need for the output capacitor.
This layout added the problems of LS
protection - most easily provided by a
simple output fuse - and the stability of
the nominally av output potential.
Various input circuit layouts have
been proposed3," to ensure that no
residual DC appeared at the loudspeak-
er output terminals, but the simplest
and most direct solution to this problem
is the use of an input long-tailed pair of
the kind shown in Fig. 12.
Provided that the emitter currents of
both devices are the same, and that they
have similar values of current gain, the
output offset will be close to zero if the
base circuit resistances for both transis­
tors are the same. A high-impedance
tail load is desirable to ensure the
integrity of signal transfer between the
two input halves.
Gain stage circuit designs
The gain stages between the signal­
input point and the output devices are
normally operated in class A and are
configured to provide as wide a band­
width, as high a gain and as Iow a phase
shift as practicable.
To simplify loop-stability problems,
Fig. 9. Symmetrical compound emitter­
follower. Bias is less temperature­
dependent.
NFB etc .
.J-t-...
--�-�----"" v
Fig. 1 O. lnternal structure of n-p-n
Darlington transistor.
Fig. 12. Long-tailed-pair input circuit
ensures that no DC is present at output.
Current mirror
Fig. 13. Current mirror presents high
dynamic-impedance load.
Current mirror
r-------....------------.
+v
....----.
Ein
--I I-....----t-t.
.------Eout
�--�----��-- -v
Fig. 11. Use of symmetrical supplies
a voids need for blocking capacitor.
1046
Fig. 14. Current-mirror shifted to second stage, as used in ICs by National
Semiconductor and by Hitachi in an audio power amplifier.
ELECTRONICS WORLD + WIRELESS WORLD
November 1989
EVOLUTIONARY AUDIO
the gain block is normally restricted to
two stages and, to get as high a gain as
possible, the collector load for the
second stage has as high a dynamic
impedance as practicable. This is often a
"bootstrapped" load resistor, as em­
ployed in the designs of Figs 5 and 6.
However, in more recent circuits, a
constant-current source load is normal­
ly used, since this gives rather better
distortion characteristics, especially at
LF, though the possible total output
voltage swing may be rather less.
The load for this input stage may just
be a single resistor, in the first collector
circuit, as shown in Fig. 12 although,
following the practice in IC op-amps, it
is more common to use a current mirror
in this position, as shown in Fig. 13.
An interesting development of this
idea is to move the current mirror to the
position of load for the second gain
stage, as shown schematically in Fig. 14.
This is an idea which appears to be due
to National Semiconductor and is em­
ployed in several of its IC op-amp
designs, such as the LH0061. This has
been adapted, more recently, to an
2
amplifer circuit by Hitachi. 1 .
Loop stability and transient
intermodulation distortion
If negative feedback is applied around a
circuit enclosing a two-stage gain block
as well as an output emitter-follower
system, it is probable that the total
phase shift within the loop will be 1800
at some frequency at which the gain is
unity, and the amplifier will oscillate.
It is essential, therefore, to ensure
stability by causing the open-loop gain
to fall as the frequency approaches the
upper (or lower) 1800 phase-shift
points. With most direct-coupled cir­
cuits, the LF loop phase shift will not
exceed a safe value ; stability problems
are therefore confined to the HF end of
the pass-band.
It was, and is, customary to achieve
Hr loop stabilization by imposing a
single-pole dominant-Iag characteristic
on the system by connecting a small
capacitor between base and collector of
the second gain stage (C2 in Figs. 12, 13
and 14), since this arrangement gives
the best THD performance at high
frequencies. However, this approach
leads to the problem that it imposes a
finite speed of response on the second
gain stage while C2 charges or dis­
charges through its associated base and
collector circuits.
If a composite signal including a step
waveform is then applied to the input
device, it is possible for the input stage
to be driven into overload because no
November 1989
I etc
1
I
.
•
Fig. J6.lnput RC filter restricts rate of input voltage change to
Fig. 15. Effect of slew-rate limiting.
that of rest of circuitry.
----.-+v
r((,
�----�--4_-- -V
Fig. 17. Preferred position for HF loop compensation capacitor.
compensating feedback signal has yet
had time to arrive from the subsequent
amplifying stages. This can lead to a
complete loss of signal during the period
in which the second gain stage is para­
lysed, and caused Otala13 to apply the
term "transient intermodulation distor­
tion" to the perceived acoustic effect.
A simpler description suggested by
Jungl4 is "slewing-induced distortion"
(or slew-rate limiting) and this defect in
the amplifier performance is clearly
visible on an oscilloscope display, with
an appropriate composite input signal,
as shown in Fig. 15.
This defect is, however, not an inevit­
able consequence of dominant-Iag com­
pensation, since there are ways of
avoiding it IS Of these the simplest is
just to introduce an RC low-pass net­
work at the beginning of the amplifier to
restrict the rate of change of input signal
voltage, as shown in Fig. 16.
A better alternative is to include the
whole of the gain stages within the
bandwidth-Iimiting system, as used, for
ELECTRONICS WORLD + WIRELESS WORLD
example, by Bailey; and as illustrated in
Fig. 17. Placing C2 in this position
avoids the possibility of input-device
overload as a consequence of the slug­
gishness of response of later stages.
Other snags
typical amplifier might, therefore,
have the kind of circuit shown in Fig. 18
(resistors R" and Rh avoid "latch-up").
A
1047
EVOLUTIONARY AUDIO
Some temperature compensation for
the output transistor forward bias can
be obtained from a suitable degree of
thermal contact between the output
devices and Try.
The stray capacitances associated
with the collector circuit of Tr7 will
impose a maximum slewing rate on a
positive-going voltage excursion. The
collector current of Tr7 must therefore
be adequate to keep this slewing rate
sufficiently high. With this point in
mind, several designers, such as
Bongiorno'6.'7 and Borbely'H, have
offered fully symmetrical amplifier cir­
cuits of the form shown in Fig. 19, so
that the maximum practicable rate of
change of signal voltage at the gain­
stage output is not limited by the final
driver-stage constant-current source
load.
However, it is more difficult to main­
tain a stable value of output-stage quies­
cent current with this type of circuit
layout, and this has discouraged its
more widespread adoption.
Fig. 19. Driver stage by Bongiorno, which does not suffer from limitation of Fig. 18
circuit.
Fig. 18. Typical fully complementary
audio power amplifier, incorporating
the features discussed. Slew rate is
limited by Tr7 collector current.
References
I. Lin, H.C., Electronics, September 1956,
pp. 173-75.
2. Locanthi, B.N., 1. Audio Eng. Soc., July
1967, pp.290-294.
3. Bailey, A.R., Wireless World, May 1968,
pp.94-98.
4. Linsley Hood, J.L., Wireless World, April
1969, pp.148-153.
5. Shaw, l.M., Wireless World, June 1969,
pp.265-266.
6. Baxandali, P.J., Wireless World, Septem­
ber 1969, pp.4 I 6-417.
7. Linsley Hood, J.L.,
Hi-Fi News and
Record Review, November 1972, pp.2120-
2123.
8. Acoustical Manufacturing Co. Ltd, Wire­
less World, April 1968, p.67.
9. Visch, N.M., Wireless World, April 1975,
p.166.
10. Stevens, G.A., Hi-Fi News and Record
Review, February 1971, pp.252-253.
11. Linsley Hood, J.L., Studio Sound, April
1975, pp.22-28.
12. Hitachi Ltd, Power Mosfet Application
Book (1981), pp.llO-ll5.
13. Otala, M., J. Audio Eng. Soc., No 6,
1972, pp.396-399.
14. Jung, W.G., Hi-Fi News and Record
Review, November 1977, pp.115-123.
15. Linsley Hood, J.L., Hi-Fi News and
Record Review, January 1978, pp.81-83.
16. Bongiorno, J., Audio, February 1974,
pp.47-51.
17.
Bongiorno,
J.,
Popular
Electronics,
September 1974, pp.33-39.
18. Borbely, E., Audio Amateur, February
1984, pp.13-24.
1048
ELECTRONICS WORLD + WIRELESS WORLD
I
November 1989
EVOLUTIONARY AUDIO
n the first part of this article, I
covered some of the develop­
ments in the design of transistor
audio power amplifiers from the
commercial introduction of transistors to about 1975, by which date
some competently engineered designs
had been produced.
A 'fair proportion of the designs pro­
duced at the end of this period were
capable of a performance which would,
to the ear of an unprejudiced listener, at
least equal most of the previous genera­
tion of valve operated equipments and
were also more compact, cooler running
and of substantially greater potential
output power.
However, design mistakes had been
made and some units having a relatively
poor acoustic performance had been
produced, particularly during the ear­
lier years of this period. Although there
was a better understanding of the re­
quirements for audio power amplifiers,
some relatively indifferent designs were
still being offered. Even in the case of
the good designs, some residual intrin­
sic problems remained.
There was the need to ensure that the
quiescent current of the output transis­
tors, in the typical class AB output
mode, was correctly set on manufacture
and remained correct during the life of
the equipment. There was also the
problem of time lag in the thermal
compensation circuitry, which could
mean that the quiescent current setting
could be in error at the onset of a burst
of high output power or in the period
immediately following it.
In addition, the relatively high
amounts of negative feedback normally
employed in these designs could cause
sporadic malfunction when used with
loudspeakers which had awkward impe­
dance characteristics, making the
amplifiers prone to "hard" clipping on
signal overload. This effect would effec­
tively require a larger transistor ampli­
fier to deliver the same amount of
apparently undistorted output power to
the speaker than would have been the
case with a valve design.
I
Design trends
At this time, three separate design
trends began to emerge, of which the
most explicable, from the engineering
point of view, was that of removing or
1164
SOLI -STATE
AUDI
POW R
John linsl e y Hood
continues his
examination of the
l u tion of transisto
audio power
amplifiers with a look
at methods of
reducing
res idu a l defects
lessening the residual effects of transis­
tor designs, such as the non-linearity of
the class AB push-pull output stage ; the
variability of, or the need to pre-set
some chosen value for, the output stage
quiescent current ; and, in earlier de­
signs, the need to use high levels of
negative feedback to achieve acceptably
low levels of harmonic distortion.
The second line of development, pur­
sued with great vigour in Japan, was
that of seeking needlessly high levels of
steady/state linearity and, in the USA,
equally unnecessary - in normal domes­
tic use - levels of output power and
bandwidth.
This technical development was
mainly spurred on by the belief of the
'man in the street' that he needed high
output powers and that large band­
widths and very low THD levels were
synonymous with perceived sound qual­
ity. The same reasoning would lead to
the argument that it was the difference
in engine capacity which made a
220B HP Mercedes a quieter and more
comfortable car than a Citroen 2CV.
Few lay enthusiasts would accept that
they could not hear any difference be­
tween two units whose only dissimilarity
was that between 0.005% THD and
0.05% THD at any point within the
audio pass-band ; or that, in the majority
of cases, their needs could probably be
comfortably met by 5W of peak audio
output power.
The third design trend was a whole­
hearted, and perhaps cynical adoption
of pseudo-scientific ideas offered by
ELECTRO NICS WORLD + WI RELESS WORLD
December 1989
EVOLUTIONARY AUDIO
+'V,
eccentric innovators on the fringes of
the 'audiophile' fraternity, particularly
when these ideas were applauded by the
quasi-technical 'hi-fi' press. The hope
was, one supposes, that equipment de­
signed in accordance with these ideas
might be applauded by the pundits and
so become the acoustic criterion by
which all other equipment would be
judged.
As an engineer, I am more in sym­
pathy with the first of these design
tren�s because their targets are clear
and their aims are explicable.
Circuit developments
Blomley. One of the first serious
attempts to overcome the difficulties of
defining and maintaining the correct
quiescent current setting for the output
transistors was that due to Blomleyl,
who proposed that crossover distortion
should be avoided by arranging that the
output transistors were biased per­
manently to a point at the beginning of
the linear part of their Vtile characteris­
tics. The preceding part of the circuit, of
which the whole is shown in schematic
form in Fig. 1, is then designed to
present the output stage with an input
signal divided into two halves by means
of a preceding switching stage, so that
the output devices are only required to
provide an output current which in­
creases from the pre-set quiescent level.
This is effectively a class B driver
stage, but the small-signal switching
stage can do this job much more accur­
ately and cleanly than the power output
devices could ever do and the small­
signal switching stage is unlikely to
suffer from thermal drift as a result of
the total power output of the amplifier.
Although the idea is sensible and
practical, no commercial unit based on
this system has been offered.
+60V
NF
T�p
LS(1Sfi)
Fig. 1. Simplified Blomley 30W ampli­
fier, with a small-signl11 switching stage
doing thejob of a class B output stage.
Fig. 2. Distortion correction by error
take-off, due to Sandman.
Fig. 3. Iterative feed-forward is theoreti­
cally able to reduce distortion as much as
required by the use of more feed­
forward stages.
Error feedforward. This method of re­
ducing system distortion was envisaged
by B1ack2, the inventor of the negative­
feedback technique, though at the time
of its invention adequate components
were not available and it was neglected.
The method was resurrected by
Sandman3 in an interesting contribution
in which he showed two practical exam­
ples of amplifiers in which distortion
was reduced by feeding forward an
error signal to the loudspeaker; these
D ecember 1989
ELECTRONICS WORLD + WIRELESS WORL[
1165
EVOLUTIONARY AUDIO
are shown schematically in Figs. 2 and 3.
In the case of the iterative feedforward
system of Fig. 3, the distortion could in
theory be reduced to as low a value as
required by the use of extra feed­
forward stages.
The other approach, applying the
error signal to the 'earthy' end of the
load, is theoretically capable of com­
pletely removing all signal errors, in­
cluding all forms of noise and waveform
distortion introduced by the main
amplifier, but will require some set-up
adjustment as well as a floating speaker
return terminal.
Current dumping. This rather inele­
gantly named circuit arrangement, in­
4
troduced by Albinson and Walker of
the Acoustical Manufacturing Com­
pany and shown in outline form in Fig.
4, appears superficially similar to Sand­
man's feed-forward circuit of Fig. 2,
except that it requires neither preset
adjustments nor a floating 'earthy' load
return point, although this similarity
Fig. 4. Acoustical Quad current-dumping amplifier, similar to the Fig. 2 Sandman
circuit except that it needs no presets or floating load.
was disputed in a subsequent letter from
Sandman5.
Of all the circuit designs so far
offered, this one seemed to come
closest to the ideal transistor layout in
that the power transistors could operate
without any forward bias whatever and
yet allow the low-distortion, low-power
amplifier to fill in the residual discon­
tinuities.
Certainly this design has excited an
enormous amount of interest from
other design engineers, if the number of
published letters and articles seeking to
explain or deny its operation is any
indication. For me, the most intellec­
tually satisfying explanation of its
method of operation is that due to
Baxanda1l6 and is as follows.
Consider a simple amplifier arrange­
ment of the kind shown in Fig. 5(a),
consisting of a high-gain linear amplifier
Al driving an unbiased pair of power
transistors Trl and Tr2 and feeding a
load ZL' Without any feedback, the
input/output transfer curve of this cir­
cuit would have the shape shown by line
(a) in Fig. 6, in which the slope would be
steep from M' to N' while Tr2 was
conducting, much flatter between N'
and N while only amplifier Al was
contributing through R3 to the load
current, and then steeper again from N
toM, while Trl was conducting.
If overall negative feedback is applied
via RI, the kink in the transfer curve can
be reduced, especially if the gain of AI is
very high, giving a more linear charac­
teristic of the type shown by line (b) in
Fig. 6. However, it would still be un­
satisfactory.
1166
Fig. S. Operation of the Quad circuit. Basic arrangement of unbiased transistors at
(a) is improved by addition of resistor R" which allows almost total elimination of
output transistor distortion.
What is required is some method of
increasing the amount of feedback
while Trl and Tr2 are conducting to
reduce the overall gain so that the slope
of the transfer characteristic M'-N' and
N-M is identical to that N'-N.
This can be achieved, as shown in Fig.
5(b), by inserting a small resistor R4
between points F and G in the output·
feed from Tr1.2 and then deriving addi­
tional feedback from point F. If the
values of R .2 are correctly chosen in
1
relation to the open-loop gain of AI>
and the output transistors TrL2 have
identical characteristics, the distortion
due to the unbiased output transistors
vanishes.
Unfortunately, resistor R4 would be
wasteful of power, so Walker and
Albinson replace it with a small induc­
tor and substitute a small capacitor for
R2 to compensate for the frequency-
dependent impedance of the inductor.
While this substitution delivers a per­
formance within the range expected
from the component tolerances, it com­
plicates the theoretical analysis of the
circuit and has led to a lot of subsequent
debate, in which the most detailed ex­
amination is that due to McLoughlin7.
He makes a number of valid objections:
that it is unlikely that the circuit will
completely remove distortion, since no
feedback amplifier can ever do this; that
the distortion 'cancellation' depends
heavily on the precision of the compo­
nents in the 'bridge' network; and that it
presumes that the output slope from M'
to N' in Fig. 6 will be identical to that
fromN toM.
Nevertheless, the circuit works and
gives a performance comparable to that
obtainable by more conventional
means, but without the need to set the
ELECTRONICS WORLD + WIRELESS WORLD
December 1989
EVOLUTIONARY AUDIO
mode) is shown in schematic form in
Fig. 7.
Fig. 6. Transfer characteristic of Fig.
Sea) circuit, with (b) and without (a)
feedback.
output transistor quiescent currents which was the initial objective.
Power mosfets. Junction transistors suf­
fer from a number of inherent prob­
lems; such as hole storage and prone­
ness to secondary breakdown and ther­
mal runaway, which becomes more con­
spicious when they are used as output
devices. With a view to avoiding these
problems. Sony introduced high-power
junction fets, suitable for use as audio
amplifier output devices, in the early
1970s and an amplifier using these was
marketed.
However, the parallel development
of the insulated-gate power mosfet
overtook that of the power fet and, by
the late 1970s, there was a range of
robust devices with greatly superior
characteristics to that of the bipolar
junction transistor. Not only are they
very fast but, if good chip geometry is
employed, the relationship between
gate voltage and drain current within
the conducting region can be very linear
indeed, which facilitates low-distortion
push-pull operation. Their very high
operating speed allows a substantial
improvement to be made in the per­
formance of a quite straightforward
audio amplifier by the mere substitution
of power mosfets for bipolar power
devices, as for example in two designs of
my ownx.
With some exceptions, circuit desig­
ners have been slow to adopt these
devices, in spite of their attractive fea­
tures.
Sandman's class S system. A very in­
teresting idea, introduced by Sandman9
and somewhat confusingly labelled
"class S" (this definition had been used
before to refer to a valve grid-bias
Dece mber 1989
This employs a high-gain error ampli­
fier A2 to sense the difference betwe€n
the output of the small-signal driver
amplifier AI and that from the unbiased
output devices Tr 1.2 to drive these so
that AI sees a very high impedance
load, under which condition its per­
formance approaches the ideal. As in
the current-dumping circuit, the input
amplifier provides a drive voltage to the
load when the power output devices are
non-conducting.
This idea has been adopted in several
Japanese power amplifiers and a simpli­
fied version of the output stage of the
Technics SE-A 100 power amplifier which is representative of all their cur­
rent range - is shown in Fig. 8.
With reference to my earlier com­
ments on the preoccupation of some
manufacturers with what appear to be
needlessly high specifications, this de­
sign is a typical example, in that it offers
a very low steady-state THD figure
(0.0002% THD at 1kHz), a very large
bandwidth (0.8Hz - 150kHz) and a high
power output (240W into a 4D load),
though with the penalty of a circuit of
considerable complexity.
Pseudo class A systems. Various other
circuit arrangements have been ex­
plored with the aim of avoiding the need
for a pre-set, and perhaps critical value
of output-stage quiescent current, with­
out the thermal and other penalties
incurred by a pure class A output stage,
such as sliding bias or other non-cut-off
layouts. Various names have been in­
vented for these, such as "class AA" or
"super A".
Of these, one of the more superficial­
ly appealing is the floating power supply
arrangement in Fig. 9.ln this layout, the
output devices Trv are operated in
class A, with a collector current which is
high enough to meet all the anticipated
output current demands of the design,
but with a supply voltage which is low
enough that the total output stage ther­
mal dissipation is within acceptable
limits.
The output-stage low-voltage power
supply is arranged to 'float', with its
+v
10k
-v
Fig. 7. Sandman's c1ass S amplifier.
Ein
-v
RL
RZ
4-efi
220
Fig. 8. Technics power amplifier output stage, using the Circuit due to Sandman.
ELECTRONICS WORLD + WIRELESS WORLD
1 167
EVOLUTIONARY AUDIO
centre tap connected to the output of a
high-power, unity-gain, class B power
amplifier. There will, of course, be
crossover-type discontinuities in the
way in which this centre-tap voltage
follows the input signal, but this will
only appear as a modulation of the
supply voltage applied to the output
transistors, and it is presumed that the
effect on the amplifier 'output will be
negligibly small.
However, there is an inherent prob­
lem, which is that the load is connected
to the OV line, but the floating power
supply is not. Since this is only returned
to this line through the class B amplifier,
it follows that this latter amplifier is in
series with the load at all times.
The system therefore relies, in prac­
tice, on the ability of the negative­
feedback loop signal to cause the pre­
ceding amplifier stages to supply a cor­
recting signal to the class A output
devices to remedy the deficiencies intro­
duced by the class B supply-line driver,
and these will only be remediable if the
class B power supply driver stage is
operated in class AB with some reme­
dial quiescent current, which must be
preset.
Also, while this system can give a
good steady-state performance, it has
problems, as have many other exotic
designs, in handling steeply rising sig­
nals, which make up so much of pro­
gramme material.
Fig. 9. Floating power-supply pseudo class A system.
Another scheme which aims to pro­
vide the advantages of class A operation
but with the economy of class AB is the
so-called 'non-switching' layout due to
Pioneer, used in their M-90 power
amplifier, for example.
The layout used is shown in Fig. 10, in
which a purpose-designed le is used to
monitor the quiescent current of each
group of output transistors and ensure
that it remains at the correct level, never
approaching cut-off. This also avoids
the need for internal pre-set adjust­
ments.
I will examine some of the remaining
aspects of this development in the con­
cluding part of this article.
•
References
1. Blomley, P. Wireless World, February 1 97 1 ,
p p . 57-6 1 ,a n d March 1 97 1 ,pp. 1 27- 1 3 1 .
2 . Black, H.S., U S Pat e nts, 1 686792 and
20 1 267 1 .
3. Sandman, A.M. Wireless World, October
1 974, pp. 367-37 1 .
4 . Walker, P.J., Wireless World, Dece mber
1 975,pp. 560-562.
5. Sandman, A.M., Wireless World,
1 976,p. 54.
",
April
6. Baxandall, P.J., Wireless World, July 1 976,
pp. 60-6 1 .
7. McLoughlin, M . , Wireless World, Septem­
ber 1 983, pp. 39-43, and October 1 983, pp.
35-4 1 .
8 . Linsley Hood, J.L., Hi-Fi News and Record
Review, January 1 980, pp. 67-7 1 , and Decem­
ber 1 980,pp. 83-85.
9. Sandman, A.M., Wireless World, Septe mber
1 982,pp. 38-39.
Fig. 10. Output transistor quiescent-current control in Pioneer M-90 (BK) amplifier.
+Vcc
LS
1
_J
1 1 68
E LECTRONICS WORLD + WIRELESS WORLD
Dece mber 1 989
EVOLUTIONARY AUDIO
it is preferable to achieve a
degree of linearity in the
sfer characteristics of the
lifier without having to use
large amounts of negative
to straighten out the kinks,
designers have paid much attention to
the design of those stages which provide
the bulk of the voltage gain within the
power amplifier.
Gain stage design
The principal techniques at the disposal
of the circuit designer in his pursuit of
greater linearity are the use of long­
tailed pair gain stages, since these tend
to lessen the generation of even-order
distortion components; the cascode
connection of the devices in the various
ways shown in Fig. 1, because this
isolates the amplifying device from the
output voltage swings; and the use of
highly symmetrical driver stage layouts,
which can lessen problems due to slew­
rate limiting. All of these methods are
exploited, in various combinations, in
contemporary circuit designs.
It is practicable to obtain high gain
with wide bandwidth simply by
cascading a series o f amplifier stages, as
in the relatively early design due to
I
Lohstroh and Otala shown in outline in
Fig. 2, but the cumulative phase errors
of succeeding stages make overall loop
stability more difficult to achieve.
Nevertheless, this approach has been
a dopted commercially; a design
employed b y Pioneer i n their M-90
power amplifier, shown schematically
in Fig. 3, shows strong similarities to the
Lohstroh/Otala layout. This Pioneer
design also shows a trend, which is
increasingly favoured in Japan, of using
cascode-connected (monolithic) dual­
junction fet inputs stages, because of
the ease of matching the DC offset
characteristics in a monolithic pair, and
the greater input linearity of fets in
comparison with bipolar transistors.
The bipolar cascode devices, Tr2 and
Tr4, which can be high-voltage working
types, then allow the supply line
voltages to be chosen without the
constraints imposed by the relatively
low gate/drain breakdown voltages of
the fets.
16
SOLI -STATE
AUD
POW R
In this fina part,
John Linsfey Hood
considers gain stages'
and power supplies,
and takes a quizzical
look at testing and
specifications.
The use of high-voltage, small-signal
mosfets in place of cascode isolated
junction fets as the input devices, as
2
adopted in a recent design of my own
shown in Fig. 4, allows a simpler layout
without loss of performance, provided
that some initial set-up adjustment is
made to compensate for possible bias­
voltage differences between the two
input devices.
The performance of the gain stage is
enhanced by cascode connecting the
driver stage preceding the output
emitter followers, as shown in the two
designs of Figs. 5 and 6 due to
.4
Borbely3 , since this stage will be
required to handle a large signal-voltage
swing.
Cascode connection, in this case,
improves the effective linearity of the
input device, particularly in respect of
collector voltage modulation of the
current gain (Early effect), and also
eliminates unwanted effects due to the
collector/base feedback capacitance.
Figure 7 shows an elaboration of this
layout used in the Technics SE-A WO
amplifier, in which the combination of
the emitter-follower group Trx.9 and the
current mirror formed by TrlO.II.15 is
used to achieve a symmetrical drive
system from a less complex single­
ended input stage, which makes it easier
to control the output stage quiescent
current· than with a fully symmetrical
driver layout, even though this may be
theoretically superior.
Although the availability of high-
ELECTRONICS WORLD + WIRELESS WORLD
January 1990
EVOLUTIONARY AUDIO
Fig.
J. Fet/bipolar cascode combina­
tions, giving good input/output isolation.
Circuit at (a) gives high gain,
high
output impedance and high-voltage op­
(a)
Eout
( b)
(c)
(d)
Eout
eration; (b) gives very high Z;II' high Zo
and high voltage; (c) very high Z;II and Zo
and low/medium voltage; (d) high gain.
very high Zo and low/medium voltage.
Vref
Fig. 2. High-quality amplifier design by
Lohstroh and Otala, giving high gai,.
and wide bandwidth by the use of several
gain stages.
Fig. 3. Pioneer's M-90 amplifier, a com­
mercial embodiment of the Fig. 2 cir('uit.
t-----�-------Eout
To output
devices
....---Eout
.
--__.---______-.____-.__________1-__________�----_t--------------�--
Vret'
DC ---+---,.
+Vcc
....----Eout
(To output
devices)
....-----Eout
---4-�----�--+-- �dd
January 1990
ELECTRONICS WORLD + WIRELESS WORLD
17
EVOLUTIONARY AUDIO
voltage devices has led to the increasing
use of linear ICs in driver gain stages,
thoses designs aimed at the upper end of
the market appear to rely almost
exclusi vely on discrete-component
circuit constructions.
An exception to this is the use, as in
the Quad 405, 5 10, 520 and 606
amplifiers, of an IC op-amp as a DC
comparator, (Fig. 8) , to ensure that the
no-signal DC voltage at the loudspeaker
output terminals remains close to the
desired zero level. This is a worthwhile
a n d i n c r e a s i ngly w i dely adopted
stratagem.
Fig. 4.
High-voltage mosfets allow a
simpler design at the expense of freedom
from setting up.
----....I-. ----------""'"1-- +Vcc
Set oftset
....---.
Eout
I
J
( To output
stages I
NFB
t-"":"--- Eout
----4�--'-------.....
...
--...._
_
-Vdd
---1--...---.-----t--.,.-.,- +Vcc
3V
T
I
( To output
stages I
"'---+-- E out
3V
Power supplies
From the point of view of the purist,
there i s no substitute for an
electronically stabilized supply as the
DC source for the power amplifier,
since this will provide rails of known
and precisely controlled potential,
largely free from noise and ripple and
having a low source impedance.
It also confers the advantage, in the
case of a power amplifier, that the
output power available can be precisely
specified and unaffected by short-term
changes in the mains supply voltage.
Instantaneous power-supply clamping
or shut-down can also be brought about
in the event of an abnormal load­
current demand or a DC-offset fault
condition at the loudspeaker output
terminals.
Such a stabilized power supply offers
many advantages, including that of
better sound quality from the power
amplifier, particularly where separate
supplies are provided for the output
devices and the preceding driver stages.
This is due to the very low source
impedance of the supply lines, which
appears to confer both a more 'solid'
bass, as well as a more precise stereo
image. Suitable designs tend to be
complex, however, as in a published
twin DC supply design of my owns.
F r o m l o w - v oltage preamp l i f i e r
supplies, stabilized supply lines derived
from IC voltage regulators are now
almost universally used but, in the case
of power amplifiers, a rigidly controlled
DC supply would not meet some
specific user requirernents.
This is because a significant part of
the market consists of enthusiasts for
rock and similar music, for whom the
physical impact of the sound is an
important part to the enjoyment of the
music. In this use , the equipment is
operated at as high a sound output level
as circumstances allow, and freedom
from noticeable clipping is a substantial
advantage.
Since many peak power demands are
o f r e l a t i v e l y b r i e f d u r a t i o n , an
unstabilized power supply, having a
relatively high off-load supply line
voltage with large-value reservo i r
capacitors, will allow the amplifier to
sound appreciably ' louder' than a
similar design with a more rigidly
controlled but lower-voltage D C
supply. T h i s i s an aspect few
manufacturers can afford t o ignore.
Fig. 5. Linear high-gain stage due to
Borbely, using symmetrical configura­
tion.
18
ELECTRO NICS WORLD + WIRELESS WORLD
January 1990
EVOLUTIONARY AUDIO
Figure. 9 shows a typical modern
power supply, with entirely separate
supplies for each channel, and very
lar g e - v a l ue r e s e r v o i r c a p a c i tors.
Clearly, the output current from such a
supply could be highly destructive of the
loudspeaker system in the event of a
c o m p o n e n t f a i lu r e a n d v a r i o u s
protection systems are used, ranging
from simple fuses in the output lines to
elaborate relay protection systems, such
as that shown in Fig. 10.
H o w e v e r , w i t h a l l of t h e s e
electromechancal components included
within the loudspeaker output line,
there remains the real possibility of
poor electrical connections through
mechanical wear or contact corrosion,
which can lead to high resistance
junctions. There is also the possibility of
rectifying effects, which are of much
greater audible significance than any
benefits thought to be conferred by
ultra-low resistance speaker cables.
....-----Eout
1
To output
emitter
followers
I
....--.....:. Eout
Amplifier testing
In an ideal world, there would be some
clearly understood and universally
agreed set of standards by which the
performance of an amplifier - OT any
other component i n the sound
reproduction chain - could b e assessed.
Some of the design errors which arose
in the early days of transistor amplifiers
disclosed inadequacies in the test
Fig. 6. A nother Borbely cascode design,
with source-followers.
...._
.,.... --....:...
.:. ..,.. ---....-t.:. -....-'.,.... --'"""_
"- .....:. - .. Vcc
"J-I.-"":"- Vref 2
Fig. 8. Output DC level correction used
by Quad in which the op-amp m aintains
the no-signal direct voltage near zero.
Fig. 7. Single-ended cascode input stage
"'--__-'-_______...:._
. __
.. ___...._
Vdd
_
by Technics m akes for ease of quiescent
current adjustment.
January 1 990
ELECTRONICS WORLD + WIRELESS WORLD
19
EVOLUTIONARY AUDIO
methods then employed. Sadly, thirty
years I'ater, we are still some way from a
complete understanding of the types of
technical specification we should seek
to meet, or of the relative acoustic
significance of the known residual
errors.
Part of this problem is due to clear
differe n ces i n their respo n s e to
instrumental evaluation between the
three groups of customers ; the classical
music devotee, the rock music
enthusiast and the relatively naive, and
musically uninterested 'man in the
street' .
I n classical music and traditional jazz
played on acoustic instruments, a direct
comparison is possible between the
sound of the original performance and
that of the reproduction, allowing for
differences in the acoustic ambience of
the settings ; the importance of residual
defects in reproduction, so far as these
are identifiable, can be quantified.
Some of the early public demon­
strations staged by G . A. Briggs of
Wharfedale and P . J. Walker of Quad,
in which live and reproduced music
were directly compared in a side-by-side
demonstration, showed that even in
those days the differences could be
surprisingly small and encouraged the
belief that the performan ce tests
employed were adequate to assure
satisfactory performance, as they could
Fig. 9. Simple unstabilized power supply
for output stages used even in high­
quality amplifiers.
PreClmp
#1+
'--+-+�-------+1-------------' OV
P reamp
#.1 -
+V
Special purpose le
....-.....
....
RL1
From
am p
10
LS
1
well have been for the equipment then
being used.
For the relatively unsophisticated
buyer of equipment, the important
factors are physical appearance, the
n umber of facilities it offers, its
apparent value for money and its
numerical performance specifications,
such as power output, bandwidth, and
steady-state harmonic and
intermodulation distortion factors .
The fact that very highly specified
power amplifiers may not sound any
better, and perhaps even worse than
systems which are less well specified,
has cast some doubt on the value of
many performance measurements . This
doubt is encouraged by the growing use
of up-market equipmen t for the
reproduction of music originating
mainly from electronic or electronically
assisted instruments - which definition
must also include the human voice,
w h e r e t h i s is a u g m e n t e d b y a
microphone and amplifier - and fed
directly on to tape.
This music is also likely to have been
e x t e n sively modified duri n g the
r e c o r d i n g p r o c e s s , s o that the
performance is heard for the first time
when the disc or tape is replayed. The
judgment of the listener will therefore
be based less upon whether the
reproduced sound is accurate than on
whether it is pleasing to the ear .
Whether it is warranted or not,
enthusiasts i n sist that there are
differences in the listener appeal of the
various available units and that these
differences may not be measurable by
a n y of the n ormally specified
performance parameters . Guidance,
when needed, must therefore be sought
elsewhere .
A wide range of periodicals exists to
cater for this need and also, perhaps, to
reinforce the belief that the respective
merits of various brands of equipment
can only be assessed by comparative
listening trials carried out by (their own)
skilled and experienced reviewers.
Clearly, the abse n ce of valid
numerical or instrumental standards for
'
defin in g subjective ampI ifier
performance is a matter of wide
concern, and various attempts have
been made to set matters straight.
To involve the ear of the listener in
the assess m e n t of performa n ce,
Colloms6 and Baxanda1l 7 almost
simultaneously proposed the
substitution of the amplifier under test
Fig.
10. Typical com m ercial speaker
protection and switch-on/off m uting cir­
cuit.
20
ELECTRONICS WORLD + WIRELESS WORLD
January 1 990
EVOLUTIONARY AUDIO
for a nominal (pjlase-corrected) straight
wire, using a circuit layout of the kind
shown in Fig. 1 1 . Perhaps predictably,
the conclusions reached by these two
authors differed, with Colloms claiming
that there were significant differences
which could be detected by this method
and Baxandall asserting t h a t all
competently designed units; operated
within their limits, will sound identical.
An early observation of audio
enthusiasts was that, in spite of their
generally poorer specifications, valve
amplifiers " sounded b e t t e r " than
transistor amplifiers. This was probably
because the valve amplifiers had a more
gradual overload characteristic than
their transistor equivalents, especially
since most solid-state amplifiers would
use o u t p u t - t ransistor protect ion
circuitry, which would impose a rigid
limit on the permissible output current
into a short circuit or low-impedance
load. Valve amplifiers did not impose
this output current limitation and for
both of these reasons could sound
significantly 'louder' than notionally
more powerful transistor operated
systems.
In an attempt to test the validity of
these claims for the audible superiority
of valve amplifiers, the Acoustical
M a n u f a c t uring Company ( Q uad)
commissioned a series of double-blind
group listening trials, reported by
Moirx, in which the panel was selected
to include people who had published
their beliefs that there were significant
differences between amplifier types and
that valve amplifiers were superior. 1n
the event, the conclusions of this trial
were that there was no statistical
significance in the group preferences,
individually or collectively, between the
Quad 303 and 405 transistor amplifiers,
or between either of these and the Quad
II operated amplifiers.
However, a possibly important factor
was that the output signals from the
amplifiers were monitored with an
oscilloscope to ensure that at no time
were Jhe output levels high enough to
cause clipping, however briefly.
As an extension to this valve versus
transistor debate, Hiraga9 tried to relate
the claimed sound differences between
the two amplifier types to test results
derived from wide-band spectrum
analysis. In general, his findings
confirmed that the listener did not
necessarily prefer undistorted signals.
A further attempt to provide a test
method to give better correlation with
the subjective assessment than simple
THD or bandwidth measurements was
evolved by the BBC and described by
January 1 990
Phose correction network
o--t----1
Good quality
power amplifier
- - - - 4>- - - -
Amplifier under test
lS
Dummy
load
Fig. 11. Circuit for "straight-wire" sub­
stitution test on audio amplifiers.
BeIcher 1 o, using weighted pseudo­
random noise signals followed by a
comb-filter rejection network.
This gave very good correlation with
a listening-panel assessment of sound
quality impairment through various
causes, which showed that the nature
and linearity of the transfer
c h a r a c t e r i s t i c o f t h e system was
i m p o r t a n t . This c o n c l u s i o n was
corroborated by Hirata 1 1 , who evolved
a test method based on an asymmetrical
pulse waveform input, in an attempt to
discover why it was possible to hear and
identify the audible defects of an
amplifier in the presence of much larger
defects introduced by the loudspeaker.
U nfortunately, the gulf between
engineers and the subjective-sound
f r a t ernity s t i l l remains, one side
claiming that any differences between
well designed amplifiers will be
vanishingly smal l , a n d the other
asserting that dramatic changes in
performance can be made by such
unlike l y actions as replacin.g the
standard mains cable with a more
expensive one.
The absurdity of some of these claims
2
provoked Self l
into a defence of
engineering st andards against the
metaphysical assertions of the 'add-on'
f r a t e r n i t y . A s I i n d i c a t e d in a
subsequent letter l \ I feel that we may
still have things to learn, outside the
comfortable realms of the steady state.
As engineers, we have made mistakes
in t h e p a s t through the lack of
stringency in the tests we applied. This
experience must make us more cautious
in claiming perfection as a result of
favourable responses to a limited
number of possibly inappropriate test
ELECTRONICS WORLD + WIRELESS WORLD
Osc illoscope
measurements; we may still have
overlooked something.
For myself, I bel ieve that some
audible differences do remain between
apparently impeccab l y specified
amplifiers, particularly where these are
based on dissimilar design philosophies
and I think some of these audible
differences are related to quite clearly
visible, and measurable, differences in
t h eir step-fu n ction response
characteristics. There are certainly
other things which also have an effect on
sound quality which we could measure,
•
if only we knew where to look.
References
I . Lohstro h , J . , and O t a l a , M . , AES 44th
Conve n t i o n , Rotterd a m , 1 973 , Ref. H6
2.
Linsley
Hood ,
J.
L. ,
Electronics and
Wireless World, March I l)Xl) , pp. 26 1 -264 .
3 . B o rbe l y , E . , Wireless World. March I l)R3 ,
p p . 69-75 .
4.
Borbely, E . , A udio A m ateur. February
1 9X4, pp. 1 3 -24 .
5.
Linslcy
Hood ,
J.
L. ,
Electronics
and
Wireless World. May I l)Xl) , pp. 524-52 7 .
6.
Colloms ,
M.,
Hi-Fi News and Re(ord
Review, Octobe r 1 l)77 , pp. X3-X5 .
7.
Baxandall ,
P.
J.,
Wire less
World.
N o v e m b e r 1 l)7 7 , p p . 63-66.
R. M o i r , J . , Wireless World, J u ly 1 l)7 R , p p .
55-5X.
9 . H i raga , J . , Hi-Fi News a n d Record Review.
March 1 l)7 7 , p p . 4 1 -45 .
1 0 . B e l c h e r , R. A . , Wireless World, May 1 l)7R ,
p p . 36-4 1 .
1 1 . H i ra t a , Y . , Wireless World, October I l)R I ,
p p . 49-52 .
12.
Self.
D. R.G. ,
Electronics and
Wireless
World, July I l)R X , p p . 6l)2-6l)6.
13.
Linsley
Hood.
J.
L.,
Electronics and
Wireless World, September I l)X R , pp. R60-R6 1 .
kelth@snookeu
21